Algorithms. Algorithms 4.3 MINIMUM SPANNING TREES. introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

Transcription

1 Algorithms ROBERT SEDGEWICK KEVIN WAYNE 4.3 MINIMUM SPANNING TREES Algorithms F O U R T H E D I T I O N ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context Last updated on 11/14/17 5:16 AM

2 4.3 MINIMUM SPANNING TREES Algorithms ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

3 Spanning tree Def. A spanning tree of G is a subgraph T that is: A tree: connected and acyclic. Spanning: includes all of the vertices. graph G spanning tree T 3

4 Spanning tree Def. A spanning tree of G is a subgraph T that is: A tree: connected and acyclic. Spanning: includes all of the vertices. not connected 4

5 Spanning tree Def. A spanning tree of G is a subgraph T that is: A tree: connected and acyclic. Spanning: includes all of the vertices. not a tree (cyclic) 5

6 Spanning tree Def. A spanning tree of G is a subgraph T that is: A tree: connected and acyclic. Spanning: includes all of the vertices. not spanning 6

7 Minimum spanning tree problem Input. Connected, undirected graph G with positive edge weights edge-weighted digraph G 7

8 Minimum spanning tree problem Input. Connected, undirected graph G with positive edge weights. Output. A spanning tree of minimum weight. Brute force. Try all spanning trees? minimum spanning tree T (weight = 50 = ) 7 8

9 Minimum spanning trees: quiz 1 Let T be a spanning tree of a connected graph G with V vertices. Which of the following statements are true? A. T contains exactly V 1 edges. B. Removing any edge from T disconnects it. C. Adding any edge to T creates a cycle. D. All of the above. spanning tree T of graph G 9

10 Image processing MST dithering 10

11 Models of nature MST of random graph 11

12 Medical image processing MST describes arrangement of nuclei in the epithelium for cancer research 12

13 Slime mold grows network just like Tokyo rail system Rules for Biologically Inspired Adaptive Network Design Atsushi Tero, 1,2 Seiji Takagi, 1 Tetsu Saigusa, 3 Kentaro Ito, 1 Dan P. Bebber, 4 Mark D. Fricker, 4 Kenji Yumiki, 5 Ryo Kobayashi, 5,6 Toshiyuki Nakagaki 1,6 * 13

14 Applications MST is fundamental problem with diverse applications. Dithering. Cluster analysis. Max bottleneck paths. Real-time face verification. LDPC codes for error correction. Image registration with Renyi entropy. Find road networks in satellite and aerial imagery. Reducing data storage in sequencing amino acids in a protein. Model locality of particle interactions in turbulent fluid flows. Autoconfig protocol for Ethernet bridging to avoid cycles in a network. Network design (communication, electrical, hydraulic, computer, road). Approximation algorithms for NP-hard problems (e.g., TSP, Steiner tree). 14

15 4.3 MINIMUM SPANNING TREES Algorithms ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

16 Simplifying assumptions For simplicity, we assume: The graph is connected. MST exists. The edge weights are distinct. MST is unique. see Exercise Note. Algorithms still work correctly even if duplicate edge weights no two edge weights are equal

17 Cut property Def. A cut in a graph is a partition of its vertices into two (nonempty) sets. Def. A crossing edge connects a vertex in one set with a vertex in the other. Cut property. Given any cut, the crossing edge of min weight is in the MST. crossing edges connect gray and white vertices minimum-weight crossing edge must be in the MST 17

18 Minimum spanning trees: quiz 2 Which is the min weight edge crossing the cut { 2, 3, 5, 6 }? A. 0 7 (0.16) B. 2 3 (0.17) C. 0 2 (0.26) D. 5 7 (0.28)

19 Cut property: correctness proof Def. A cut is a partition of a graph s vertices into two (nonempty) sets. Def. A crossing edge connects two vertices in different sets. Cut property. Given any cut, the min-weight crossing edge e is in the MST. Pf. Suppose e is not in the MST. Adding e to the MST creates a cycle. Some other edge f in cycle must be a crossing edge. Removing f and adding e is also a spanning tree. Since weight of e is less than the weight of f, that spanning tree has lower weight. Contradiction. f e the MST does not contain e adding e to MST creates a cycle 19

20 4.3 MINIMUM SPANNING TREES Algorithms ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

21 Weighted edge API Edge abstraction needed for weighted edges. public class Edge implements Comparable<Edge> Edge(int v, int w, double weight) create a weighted edge v-w int either() either endpoint int other(int v) the endpoint that's not v int compareto(edge that) compare this edge to that edge double weight() the weight String tostring() string representation weight v w Idiom for processing an edge e: int v = e.either(), w = e.other(v); 21

22 Weighted edge: Java implementation public class Edge implements Comparable<Edge> { private final int v, w; private final double weight; } public Edge(int v, int w, double weight) { this.v = v; this.w = w; this.weight = weight; } public int either() { return v; } public int other(int vertex) { if (vertex == v) return w; else return v; } public int compareto(edge that) { if (this.weight < that.weight) return -1; else if (this.weight > that.weight) return +1; else return 0; } constructor either endpoint other endpoint compare edges by weight 22

23 Edge-weighted graph API public class EdgeWeightedGraph EdgeWeightedGraph(int V) EdgeWeightedGraph(In in) create an empty graph with V vertices create a graph from input stream void addedge(edge e) add weighted edge e to this graph Iterable<Edge> adj(int v) edges incident to v Iterable<Edge> edges() all edges in this graph int V() int E() number of vertices number of edges String tostring() string representation Conventions. Allow self-loops and parallel edges. 23

24 Edge-weighted graph: adjacency-lists representation Maintain vertex-indexed array of Edge lists. V tinyewg.txt 8 E adj[] Bag objects references to the same Edge object Edge-weighted graph representation 24

25 Edge-weighted graph: adjacency-lists implementation public class EdgeWeightedGraph { private final int V; private final Bag<Edge>[] adj; public EdgeWeightedGraph(int V) { } this.v = V; adj = (Bag<Edge>[]) new Bag[V]; for (int v = 0; v < V; v++) adj[v] = new Bag<Edge>(); public void addedge(edge e) { } int v = e.either(), w = e.other(v); adj[v].add(e); adj[w].add(e); same as Graph, but adjacency lists of Edges instead of integers constructor add edge to both adjacency lists } public Iterable<Edge> adj(int v) { return adj[v]; } 25

26 Minimum spanning tree API Q. How to represent the MST? public class MST MST(EdgeWeightedGraph G) constructor Iterable<Edge> edges() edges in MST double weight() weight of MST 26

27 4.3 MINIMUM SPANNING TREES Algorithms ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

28 Kruskal s algorithm demo Consider edges in ascending order of weight. Add next edge to tree T unless doing so would create a cycle. graph edges sorted by weight an edge-weighted graph

29 Kruskal s algorithm: visualization 29

30 Kruskal s algorithm: correctness proof Proposition. [Kruskal 1956] Kruskal s algorithm computes the MST. Pf. [Case 1] Kruskal s algorithm adds edge e = v w to T. Vertices v and w are in different connected components of T. Cut = set of vertices connected to v in T. By construction of cut, no edge crossing cut is in T. No edge crossing cut has lower weight. Why? Cut property edge e is in the MST. add edge to tree w v 30

31 Kruskal s algorithm: correctness proof Proposition. [Kruskal 1956] Kruskal s algorithm computes the MST. Pf. [Case 2] Kruskal s algorithm discards edge e = v w. From Case 1, all edges in T are in the MST. The MST can t contain a cycle. adding edge to tree would create a cycle w v 31

32 Kruskal s algorithm: implementation challenge Challenge. Would adding edge v w to tree T create a cycle? If not, add it. How difficult to implement? A. 1 B. log V C. V D. E + V add edge to tree adding edge to tree would create a cycle Case 1: v and w in same component Case 2: v and w in different components 32

33 Kruskal s algorithm: implementation challenge Challenge. Would adding edge v w to tree T create a cycle? If not, add it. Efficient solution. Use the union find data structure. Maintain a set for each connected component in T. If v and w are in same set, then adding v w would create a cycle. To add v w to T, merge sets containing v and w. w v w v Case 2: adding v w creates a cycle Case 1: add v w to T and merge sets containing v and w 33

34 Kruskal s algorithm: Java implementation public class KruskalMST { private Queue<Edge> mst = new Queue<Edge>(); public KruskalMST(EdgeWeightedGraph G) { DirectedEdge[] edges = G.edges(); Arrays.sort(edges); UF uf = new UF(G.V()); edges in the MST sort edges by weight maintain connected components } for (int i = 0; i < G.E(); i++) { Edge e = edges[i]; int v = e.either(), w = e.other(v); if (uf.find(v)!= uf.find(w)) { uf.union(v, w); mst.enqueue(e); } } greedily add edges to MST edge v w does not create cycle merge connected components add edge e to MST } public Iterable<Edge> edges() { return mst; } 34

35 Kruskal s algorithm: running time Proposition. Kruskal s algorithm computes MST in time proportional to E log E (in the worst case). Pf. operation frequency time per op SORT 1 E log E UNION V 1 log V FIND 2 E log V using weighted quick union 35

36 Greed is good Gordon Gecko (Michael Douglas) evangelizing the importance of greed (in algorithm design?) Wall Street (1986) 36

37 MAXIMUM SPANNING TREE Problem. Given an undirected graph G with positive edge weights, find a spanning tree that maximizes the sum of the edge weights. Running time. E log E (or better) maximum spanning tree T (weight = 104) 37

38 4.3 MINIMUM SPANNING TREES Algorithms ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

39 Prim s algorithm demo Start with vertex 0 and greedily grow tree T. Add to T the min weight edge with exactly one endpoint in T. Repeat until V 1 edges an edge-weighted graph

40 Prim s algorithm: visualization 40

41 Prim s algorithm: proof of correctness Proposition. [Jarník 1930, Dijkstra 1957, Prim 1959] Prim s algorithm computes the MST. Pf. Let e = min weight edge with exactly one endpoint in T. Cut = set of vertices in T. No crossing edge is in T. No crossing edge has lower weight. Cut property edge e is in the MST. edge e = 7-5 added to tree 41

42 Prim s algorithm: implementation challenge Challenge. Find the min weight edge with exactly one endpoint in T. How difficult to implement? A. l B. log E C. V D. E 1-7 is min weight edge with exactly one endpoint in T priority queue of crossing edges

43 Prim s algorithm: lazy implementation Challenge. Find the min weight edge with exactly one endpoint in T. Lazy solution. Maintain a PQ of edges with (at least) one endpoint in T. Key = edge; priority = weight of edge. DELETE-MIN to determine next edge e = v w to add to T. If both endpoints v and w are marked (both in T), disregard. Otherwise, let w be the unmarked vertex (not in T): add e to T and mark w add to PQ any edge incident to w (assuming other endpoint not in T) 1-7 is min weight edge with exactly one endpoint in T priority queue of crossing edges

44 Prim s algorithm: lazy implementation demo Start with vertex 0 and greedily grow tree T. Add to T the min weight edge with exactly one endpoint in T. Repeat until V 1 edges an edge-weighted graph

45 Prim s algorithm: lazy implementation public class LazyPrimMST { private boolean[] marked; private Queue<Edge> mst; private MinPQ<Edge> pq; // MST vertices // MST edges // PQ of edges } } public LazyPrimMST(WeightedGraph G) { pq = new MinPQ<Edge>(); mst = new Queue<Edge>(); marked = new boolean[g.v()]; visit(g, 0); while (!pq.isempty() && mst.size() < G.V() - 1) { } Edge e = pq.delmin(); int v = e.either(), w = e.other(v); if (marked[v] && marked[w]) continue; mst.enqueue(e); if (!marked[v]) visit(g, v); if (!marked[w]) visit(g, w); assume G is connected repeatedly delete the min weight edge e = v w from PQ ignore if both endpoints in T add edge e to tree add either v or w to tree 45

46 Prim s algorithm: lazy implementation private void visit(weightedgraph G, int v) { } marked[v] = true; for (Edge e : G.adj(v)) if (!marked[e.other(v)]) pq.insert(e); public Iterable<Edge> mst() { return mst; } add v to T for each edge e = v w, add to PQ if w not already in T 46

47 Lazy Prim s algorithm: running time Proposition. Lazy Prim s algorithm computes the MST in time proportional to E log E and extra space proportional to E (in the worst case). Pf. minor defect operation frequency binary heap DELETE-MIN E log E INSERT E log E 47

48 Prim s algorithm: eager implementation Challenge. Find min weight edge with exactly one endpoint in T. Observation. For each vertex v, need only lightest edge connecting v to T. MST includes at most one edge connecting v to T. Why? If MST includes such an edge, it must take lightest such edge. Why?

49 Prim s algorithm: eager implementation demo Start with vertex 0 and greedily grow tree T. Add to T the min weight edge with exactly one endpoint in T. Repeat until V 1 edges an edge-weighted graph

50 Prim s algorithm: eager implementation demo Start with vertex 0 and greedily grow tree T. Add to T the min weight edge with exactly one endpoint in T. Repeat until V 1 edges. v edgeto[] distto[] MST edges

51 Prim s algorithm: eager implementation Challenge. Find min weight edge with exactly one endpoint in T. Eager solution. Maintain a PQ of vertices connected by an edge to T, where priority of vertex v = weight of lightest edge connecting v to T. Delete min vertex v; add its associated edge e = v w to T. Update PQ by considering all edges e = v x incident to v ignore if x is already in T add x to PQ if not already on it PQ has at most one entry per vertex decrease priority of x if v x becomes lightest edge connecting x to T red: on PQ black: on MST 51

52 Indexed priority queue Associate an index between 0 and n 1 with each key in a priority queue. Insert a key associated with a given index. Delete a minimum key and return associated index. Decrease the key associated with a given index. for Prim s algorithm, n = V and index = vertex. public class IndexMinPQ<Key extends Comparable<Key>> IndexMinPQ(int n) create indexed PQ with indices 0, 1,, n 1 void insert(int i, Key key) associate key with index i int delmin() remove a minimal key and return its associated index void decreasekey(int i, Key key) decrease the key associated with index i boolean contains(int i) is i an index on the priority queue? boolean isempty() is the priority queue empty? int size() number of keys in the priority queue 52

53 Indexed priority queue: implementation Binary heap implementation. [see Section 2.4 of textbook] Start with same code as MinPQ. Maintain parallel arrays so that: keys[i] is the priority of vertex i qp[i] is the heap position of vertex i pq[i] is the index of the key in heap position i Use swim(qp[i]) to implement decreasekey(i, key). decrease key of vertex 2 to C i keys[i] A S O R T I N G - qp[i] pq[i] N 1 A 2 3 G vertex 2 is at heap index 4 8 R O S I T

54 Prim s algorithm: which priority queue? Depends on PQ implementation: V INSERT, V DELETE-MIN, E DECREASE-KEY. PQ implementation INSERT INSERT-MIN DECREASE-KEY total unordered array 1 V 1 V 2 Bottom line. binary heap log V log V log V E log V d-way heap log d V d log d V log d V E log E/V V Fibonacci heap 1 log V 1 E + V log V amortized Array implementation optimal for complete graphs. Binary heap much faster for sparse graphs. 4-way heap worth the trouble in performance-critical situations. Fibonacci heap best in theory, but not worth implementing. 54

55 4.3 MINIMUM SPANNING TREES Algorithms ROBERT SEDGEWICK KEVIN WAYNE introduction cut property edge-weighted graph API Kruskal s algorithm Prim s algorithm context

56 Does a linear-time MST algorithm exist? year worst case discovered by 1975 E log log V Yao 1976 E log log V Cheriton Tarjan 1984 E log* V, E + V log V Fredman Tarjan 1986 E log (log* V) Gabow Galil Spencer Tarjan 1997 E α(v) log α(v) Chazelle 2000 E α(v) Chazelle 2002 optimal Pettie Ramachandran 20xx E??? deterministic compare-based MST algorithms Remark. Linear-time randomized MST algorithm (Karger Klein Tarjan). 56

57 Euclidean MST Given n points in the plane, find MST connecting them, where the distances between point pairs are their Euclidean distances. Brute force. Compute ~ n 2 / 2 distances and run Prim s algorithm. Ingenuity. Exploit geometry; n log n using Delaunay triangulation. 57

58 MINIMUM BOTTLENECK SPANNING TREE Problem. Given an edge-weighted graph G, find a spanning tree that minimizes the maximum weight of its edges. Running time. E log E (or better) Note: not necessarily a MST 21 minimum bottleneck spanning tree T (bottleneck = 9) 58